Abstract:

In acoustic tomography observables like travel time can be expressed in terms of integrals along geometrical optics ray paths where the integrand is called the weighting function. In internal-wave tomography, the travel-time variance is an important acoustic observable and in this case a well-known weighting function has been derived which has the form F(z[inf ray],(theta)[inf ray])=<(mu)[sup 2](z)>L[inf p]((theta),z)/c(z)[sup 2] where <(mu)[sup 2](z)> is the fractional-sound-speed variance, L[inf p]((theta),z) is the effective internal-wave correlation length along a ray-path tangent with angle (theta), and c(z) is sound speed. Direct numerical solution for the travel-time variance weighting function shows that acoustic signals are most sensitive to internal waves near the ray upper apex but not as strongly as suggested by F. Examples comparing the direct calculation of the weighting function with F, suggest that the greatest sensitivity to internal waves exists within hundreds rather than tens of meters vertically of the ray upper apex. The inaccuracy of the weighting function, F, can be attributed to the assumptions of small ray curvature and the depth dependence of <(mu)[sup 2](z)>. Comparisons between observations and predictions of travel-time variance for basin-scale acoustic transmissions based on the new weighting function show excellent agreement for internal waves at half the Garrett--Munk spectral level.